Metabolomic markers of healthy dietary patterns in US postmenopausal women

Abstract

Healthy diet patterns are associated with lower risk of cancer and other chronic diseases. Metabolomics has the potential to expand dietary biomarker development to include dietary patterns, which may provide a complement or alternative to self-reported diet. This study examined the correlation of serum untargeted metabolomic markers with 4 diet pattern scores-the alternate Mediterranean diet score (aMED), alternate Healthy Eating Index (AHEI)-2010, the Dietary Approaches to Stop Hypertension (DASH) diet, and the Healthy Eating Index (HEI)-2015-and used multivariate methods to identify discriminatory metabolites for each pattern. Among 1367 US postmenopausal women with serum metabolomic data in the Cancer Prevention Study-II Nutrition Cohort, we conducted partial correlation analysis, adjusted for demographic and lifestyle variables, to examine cross-sectional correlations between serum metabolomic markers and healthy diet pattern scores. In a randomly selected "training" set (50%), we conducted orthogonal partial least-squares discriminant analysis to identify metabolites that discriminated the top from bottom diet score quintiles. Combinations of metabolites with a variable importance in projection (VIP) score ≥2.5 were tested for predictability in the "testing" set based on the use of receiver operating characteristic curves. Out of 1186 metabolites, 32 unique metabolites were considered discriminatory based on a VIP score ≥2.5 in the training dataset with some overlap across scores (aMED=16; AHEI=17; DASH=13; HEI=12). Spearman partial correlation analyses, applying a cut-point (|r|≥0.15) and Bonferroni correction (P<1.05×10-5), identified similar key metabolites. The top 5 metabolites for each pattern mostly distinguished high compared with low scores; 4 of the 5 (fish-derived) metabolites were the same for aMED and AHEI, 2 of which were identified for HEI; 4 DASH metabolites were unique. Metabolomic methods that used a split-sample approach identified potential biomarkers for 4 healthy diet patterns. Similar metabolites across scores reflect fish consumption in healthy dietary patterns. These findings should be replicated in independent populations.